diff --git a/docs/HARDWARE_BOM.md b/docs/HARDWARE_BOM.md index a048e36..280ef51 100644 --- a/docs/HARDWARE_BOM.md +++ b/docs/HARDWARE_BOM.md @@ -12,14 +12,14 @@ Prices and stock below are a point-in-time reference only and must be rechecked |---|---:|---|---|---|---|---:|---|---| | H1 | 1 | Waveshare ESP32-S3 2.41 inch AMOLED Touch Display Dev Board **with case** | `ESP32-S3-Touch-AMOLED-2.41-B` | The Pi Hut | `WAV-30589` | **£51.90 inc VAT** | https://thepihut.com/collections/waveshare/products/esp32-s3-2-41-amoled-touch-display-dev-board-with-case-600x450 | **Selected**. 600 x 450 AMOLED, 800 cd/m2 stated brightness, touch, ESP32-S3R8, 8 MB PSRAM, 16 MB flash, QMI8658 IMU, RTC, microSD/TF and supplied protective case. | | H2 | 1 | SparkFun GPS Breakout, u-blox NEO-M9N, **SMA** (Qwiic) | `GPS-17285` / manufacturer part `17285` | DigiKey UK | `1568-17285-ND` | **£57.99 ex VAT / £69.59 inc VAT** | https://www.digikey.co.uk/en/products/detail/sparkfun-electronics/GPS-17285/13561758 | **Selected**. Genuine NEO-M9N breakout, 25 Hz max navigation rate, 3.3 V logic, UART/I2C, UBX support and integrated SMA antenna connector. | -| H3 | 1 | Active GNSS antenna, SMA male | To be selected | UK supplier required | TBD | TBD | TBD | **Required before vehicle testing**. Must be appropriate for the M9N L1 GNSS bands, active-antenna bias and external automotive mounting. | +| H3 | 1 | Taoglas Magma X active GNSS antenna, magnetic mount, 3 m RG-174, SMA(M) | `AA.170.301111` | RS Components UK | `2857133` | **£22.15 ex VAT / £26.58 inc VAT** | https://uk.rs-online.com/web/p/gps-antennas/2857133 | **Selected**. Covers all four constellations the M9N tracks concurrently, including BeiDou at 1561 MHz, which the common 1575-1610 MHz antennas miss. 1.8-5.5 V suits the breakout's 3.3 V bias; 26-32 dB LNA; IP67; automotive-grade manufacture. | | H4 | 1 | High-endurance microSD | 8 to 32 GB | UK supplier | TBD | TBD | TBD | Session trace and event logging. FAT32 during development. | | H5 | 1 | USB-C data/power cable | quality short cable | UK supplier | TBD | TBD | TBD | Programming and bench power. | | H6 | 1 | 5 V vehicle USB supply | fused, good-quality automotive adaptor | UK supplier | TBD | TBD | TBD | Prototype vehicle power path. Do not connect raw vehicle 12 V to the Waveshare board. | | H7 | 1 | Rear GNSS enclosure/pod | 3D printed prototype | UK fabrication / in-house | custom | TBD | n/a | Holds the GPS-17285 securely and exposes/protects the SMA antenna connection. | | H8 | as needed | M2/M3 fasteners, threaded inserts and spacers | stainless/brass | UK supplier | TBD | TBD | TBD | Vibration-resistant assembly. | -**Known selected-electronics subtotal:** **£121.49 inc VAT** for H1 + H2 at the checked single-unit prices. This excludes antenna, microSD, power, enclosure fabrication and delivery. +**Known selected-electronics subtotal:** **£148.07 inc VAT** for H1 + H2 + H3 at the checked single-unit prices. This excludes microSD, power, enclosure fabrication and delivery. ### Procurement notes @@ -102,6 +102,103 @@ Key characteristics relevant to this project: The firmware acceptance requirement remains **20 Hz minimum**. Development should establish reliable 20 Hz operation first, then evaluate 25 Hz with the display, SD logging and all normal tasks running concurrently. +## D. GNSS antenna + +### Selected: Taoglas Magma X `AA.170.301111`, RS Components `2857133`, £26.58 inc VAT + +Chosen 2026-08-21, filling the last unspecified item in the reception chain. + +### Band coverage is what decides this, and it is easy to get wrong + +The M9N tracks four constellations concurrently, and they do not share a frequency: + +| Constellation | Frequency | +|---|---:| +| GPS and Galileo | 1575.42 MHz | +| **BeiDou B1** | **1561.098 MHz** | +| GLONASS L1 | 1602 +/- 8 MHz | + +Many antennas sold as "GPS/GNSS" are tuned **1575-1610 MHz**, which covers GPS, Galileo and +GLONASS but sits **above BeiDou**. SparkFun's own `GPS-14986` at around GBP 18 is one of them. +Fitting one costs a quarter of the receiver's constellations, and it does so **silently**: +nothing reports an error, there are simply fewer satellites and worse geometry, which is +exactly what hurts at a circuit surrounded by grandstands and pit buildings. + +The AA.170 covers all three bands explicitly. + +### Electrical fit + +- **1.8 V min, 3.0 V typ, 5.5 V max** - works on the `GPS-17285`'s 3.3 V antenna bias +- **26-32 dB LNA** - covers the roughly 3 dB loss of its 3 m RG-174 without over-driving the + M9N front end +- **SMA(M)** - mates directly with the breakout's SMA, with no adaptor and no U.FL pigtail +- **IP67**, manufactured to IATF 16949 + +### Mounting matters more than the antenna does + +- **It needs a ground plane**, which is what the magnet mount uses the steel roof for. Mount it + on the roof, centred, with a clear view of the sky. +- **Do not mount it inside the cabin.** An antenna on the dash under the windscreen loses + signal to the glass and gains multipath from the screen and A-pillars, which is the error + this whole selection exists to reduce. +- **Retention at track speed.** These magnets are strong and the part is used on commercial + vehicles, but an antenna leaving a car at 120 mph is a hazard to everyone else on circuit. + Fit a safety tether, or use the **AA.171**, the same antenna in an adhesive and screw-mount + body, once the pod is a permanent fit. + +### Considered and rejected + +| Candidate | Coverage | UK price inc VAT | Outcome | +|---|---|---:|---| +| **Taoglas AA.170** | 1561 / 1575 / 1602 | **£26.58** | **Selected** | +| SparkFun `GPS-14986` | 1575-1610 | ~£18 | Rejected: misses BeiDou | +| SparkFun `GPS-15192` / u-blox ANN-MB-00 | 1559-1606 and 1197-1249 | £98.84 | Rejected: L2 band unusable on an L1 receiver, four times the price | + +### Alternatives considered and rejected, 2026-08-21 + +Re-examined against the possibility of better reception. The M9N was retained. Recorded here +so the question is not reopened from scratch. + +| Candidate | Bands | Max navigation rate | Protocol | UK price inc VAT | Outcome | +|---|---|---|---|---|---| +| **NEO-M9N `GPS-17285`** | L1 | **25 Hz, four constellations** | UBX | **£69.59** | **Retained** | +| NEO-F10N | L1 + L5 | 10 Hz | UBX | ~£60 | Rejected: rate | +| SparkFun LG290P quad-band RTK | quad | 20 Hz | NMEA 0183 / RTCM | £182.40 | Rejected: protocol, RTK | +| u-blox ZED-F9P `GPS-16481` | L1 + L2 | 9 to 25 Hz, configuration-dependent | UBX | ~£205 | Not proven to meet the rate requirement | + +**NEO-F10N** is dual-band and inexpensive, and fails outright at **10 Hz** against a 20 Hz +minimum. + +**LG290P** fails on two counts. It speaks NMEA 0183 and RTCM but **not UBX**, and standard NMEA +carries time to 10 ms where UBX carries it to the nanosecond - the crossing-time interpolation +in `crossing_time.cpp` exists specifically to avoid quantising a lap to the update grid, and a +10 ms timestamp reintroduces a large part of what it removes. Separately, its headline feature +is RTK, which this project lists as an explicit non-goal and could not use anyway: RTK needs a +correction stream, and the device must work fully offline with no dependency on Wi-Fi or a +phone. + +**ZED-F9P** is the only genuinely tempting one: multi-band L1/L2, native UBX, and real +multipath rejection, which is the dominant position error at a circuit. It is **not confirmed +to meet the rate requirement**. Its data sheet quotes the maximum navigation rate as a table of +six values by constellation configuration - PVT 9 / 10 / 20 / 20 / 16 / 25 Hz, RTK 7 / 10 / 15 / +14 / 13 / 20 Hz - and the pattern indicates that the higher rates come from running fewer +constellations. If 25 Hz means GPS alone, then at full four-constellation concurrency the rate +may fall below the 20 Hz minimum, and the concurrency being given up is itself a reception +benefit. **Anyone revisiting this must confirm which configuration yields which rate before +ordering.** + +## The open reception decision is the antenna, not the receiver + +Line H3 is still "to be selected", and it is now the highest-value reception decision left. +Multipath dominates position error at circuits, and antenna quality, ground plane and mounting +position dominate multipath. A good active L1 antenna properly mounted will improve reception +more per pound than any receiver in the table above, and a poor one would waste a £205 module +entirely. + +If trace logging later shows multipath limiting lap repeatability, the ZED-F9P or the +dead-reckoning ZED-F9R become evidence-backed upgrades rather than guesses - the logging needed +to make that judgement already exists. + ### Why the SMA variant is now the baseline The previously proposed `GPS-15712` U.FL breakout has been superseded in this project by `GPS-17285`.